You are on page 1of 12

157 The Journal of The South African Institute of Mining and Metallurgy MAY/JUNE 2000

Introduction
Elandsrand Gold Mining Company Limited was
incorporated in 1974 to mine the Ventersdorp
Contact Reef (VCR) on a mining lease of 2619
hectares straddling the boundary between
Northwest Province and Gauteng. The location
of the mine is shown in Figure 1. Production
commenced in December 1978, and is
scheduled to continue well into the 21st
century from the current lease. The sub-
vertical shafts were completed in June 1984,
with final deepening being completed in 1997.
The mine is now able to access the VCR over
the entire lease area. During 1997, Elandsrand
acquired the entire share capital of Deelkraal
Gold Mining Company Limited, the
neighbouring mine on the western boundary
of the mining lease (Figure 1). On 29 June
1998, Elandsrand, Deelkraal and Western
Deep Levels were all incorporated into
AngloGold, and the three mines now form the
AngloGold West Wits Operations.
Deelkraal and Western Deep Levels will not
be included in the forthcoming discussion
except for a comparison of the seismic
response to longwall mining and sequential
grid miningbecause neither played a part in
the development of sequential grid mining.
This paper reviews the sequential grid mining
method (SGM) and its application on
A review of the sequential grid mining
method employed at Elandsrand Gold
Mine
by M.F. Handley*, J.A.J. de Lange**, F. Essrich

, and
J.A. Banning

Synopsis
Sequential grid mining has been employed as the primary mining
strategy at the AngloGold West Wits Elandsrand Mine since the
inception of stoping operations from the sub-vertical shaft system
in 1988. Little has been published on sequential grid mining as a
mining strategy since, and the purpose of this paper is to describe
sequential grid mining, its history, its implementation at
Elandsrand, and the outcome to date.
The results obtained using sequential grid mining have been
evaluated from seismic data, productivity statistics, and rock
related accident statistics. Comparisons with neighbouring mines,
namely Western Deep Levels and Deelkraal, have been made to
substantiate the improved conditions observed at Elandsrand, and
to compare sequential grid mining directly with longwall mining in
similar conditions and at similar depths. Sequential grid mMining is
shown to be cost-effective, more flexible, and safer than longwall
mining in similar conditions, at least on the Ventersdorp Contact
Reef in the western part of the Carletonville Goldfield. More work is
necessary to establish it as the favoured mining strategy at depths
exceeding 3000 metres, and for reefs other than the Ventersdorp
Contact Reef.
* University of Pretoria, Pretoria, e-mail:
mhandley@postino.up.ac.za.
** Elandsrand Mine, e-mail:
jdelange@anglogold.com.

Great Noligwa Mine, e-mail:


essrich

f@vrreng.anglogold.com.

AngloGold Training and Development Services, e-


mail: tbanning@anglogold.com.
The South African Institute of Mining and
Metallurgy, 2000. SA ISSN 0038223X/3.00 +
0.00. Paper received Dec. 1998; revised paper
received Jan. 2000.
Figure 1Location of Elandsrand Gold Mining
Company Limited
A review of the sequential grid mining method employed at Elandsrand Gold Mine
Elandsrand with the objective of describing its merits as a
deep level mining strategy. In order to do this, a description
of the geological setting of the mine is presented, followed by
the development history of sequential grid mining, including
some details of the design and rock mechanics modelling
results. Finally, an assessment of the success of SGM at
Elandsrand is given.
Geological setting of Elandsrand
The surface geology of Elandsrand consists of quartzites and
shales belonging to the Timeball Hill Formation, which forms
the west-south-westerly trending range of hills known as the
Westwitsrante. A diabase sill is present near the base of the
quartzites. Andesitic lavas belonging to the Hekpoort
Formation cover the southern part of the Elandsrand mining
lease area. The Malmani Subgroup Dolomites, the Black Reef
Formation, and the lavas of the Ventersdorp Supergroup
(Figure 2) underlie the surface rocks. The Ventersdorp
Contact Reef is found at the base of the Ventersdorp
Supergroup Lavas, about 1700 metres below surface at the
northern boundary of the Elandsrand mining lease area, and
about 3300 metres below surface at the southern boundary
of the lease.
The Ventersdorp Contact Reef is a gold-bearing quartz
pebble conglomerate capping the topmost angular
unconformity of the Witwatersrand Supergroup (Figure 2).
Regionally, it is seen as a separate lithological unit in its own
right, and is officially referred to as the Venterspost
Conglomerate Formation (Engelbrecht
1
). On Elandsrand it
strikes on a bearing of 62 east of north and has an average
dip of 24 southwards. It suboutcrops against the base of the
Black Reef Formation in a direction parallel to strike along
the northern boundary of the lease area.
The VCR exhibits highly erratic morphological and grade
characteristics, varying from zero to 3 metres thick. The
topography of the VCR depositional area was uneven, and
consisted of a series of slopes and horizontal terraces at
different elevations. The thickness of the reef is closely
related to the nature of the terrain on which it was laid down
i.e. slope reef tends to be thin and poor in gold content while
terrace reef tends to be thick, containing higher gold concen-
trations. Terrace reef therefore presents more favourable
mining conditions than slope reef.
The VCR unconformably overlies the footwall strata upon
which it rests, on average by an angle of 4 to the east. It
therefore rests on progressively older Witwatersrand
Supergroup strata eastwards, beginning with the Mondeor
Conglomerate Formation in the west, and ending with the
Elsburg Quartzite Formation in the east. The lavas of the
Alberton Formation throughout the lease area conformably
overlie the VCR. At present the VCR has been accessed and
mined from depths between 1700 and 2700 metres below
surface (approximately 3.2 kilometres on dip) within the
lease area, and over a distance of 5.5 kilometres on strike.
Faulting on the VCR at Elandsrand is relatively minor
when compared with other gold mines elsewhere in the
Witwatersrand Basin. The main structural trend lies in a
NNE-SSW direction, and a minor trend lies parallel to the
strike of the VCR. Dykes and sills ranging from
Klipriviersberg to Karoo age also intrude the strata. Many
dykes followed pre-existing fault surfaces when they
intruded, and now have throws, which previously belonged
to the faults. The throw displacements are predominantly
normal, seldom exceeding 20 metres, and nearly always 10
metres or less. Fault throws in this size range make it
possible to access the reef in a regular grid pattern, which
provides a good base for sequential grid mining within a
regular stablizing pillar layout.
The deepest stopes on Elandsrand are currently about
2700 metres below surface. The VCR is known to extend
down-dip of the current stoping, to about 3300 metres below
surface on the southern boundary of the mining lease area.
Elandsrand is therefore classified as a deep level gold mine,
and needs to employ deep level stoping methods to extract
the reef safely. Sequential grid mining with dip stabilizing
pillars provides one possible solution in the above-described
scenario.
Sequential grid mining method
The shallowest parts of Elandsrand (from 60 to 76 Levels,
17002200 metres below surface) were exploited using the
standard scattered mining method. At these depths mining
induced seismicity and rockbursting were already a
worsening problem, especially in the remnant areas between
stopes, and on geological features such as faults and dykes

158 MAY/JUNE 2000 The Journal of The South African Institute of Mining and Metallurgy
Figure 2Generalized stratigraphic column for the Carletonville
Goldfield
near the stopes. It was therefore apparent that some strategy
had to be adopted to reduce the number of remnants being
formed, while at the same time minimizing the seismicity
generated on faults and dykes.
History of the Idea
Cook et al.
2
had shown, with the newly introduced concept of
Energy Release Rate (ERR), the value of leaving reef pillars
in situ as a means of rockburst control. Their diagram to
illustrate the point has been converted into SI units, and
reproduced in Figure 3. In both the cases studied, i.e. mining
breast panels on strike and panels up-dip, reef pillars were
left on dip. This study showed that ERR was better controlled
by mining up-dip, and that a larger percentage of reef could
be extracted in this way at lower average ERR, when
compared to mining panels on strike. On the other hand,
updip mining has the practical disadvantage of relatively
small face length per longwall established, whilst breast
mining has good face length but soon encounters
unacceptably high ERR at relatively low extraction ratios (see
Figure 3).
Since good face length and flexibility in a variable
orebody would be of prime importance, a breast mining
system with panels advancing along strike is preferable. The
rising ERR predicted by Figure 3 for breast panels mining on
strike could be controlled by limiting stope span, for example
by mining on only one side of a raise at a time.
Murie
3
applied these ideas to an area east of the shaft
between 70 and 73 Levels. He found that dip rib pillars, in
combination with bracket pillars on dykes and faults, would
reduce the ERR by 50 per cent when compared with the
situation where only bracket pillars were left on the
geological features. It was clear that a systematic set of dip
pillars would do a lot to control the ERR, with the additional
advantage that mining would stop on the planned pillar
positions when ERR was approaching an unacceptably high
value. The alternative of a longwall mining system with
strike stabilizing pillars would be equally good from an ERR
point of view, but would be inappropriate because of the
variability of grade in the VCR. Also, a significant proportion
of seismicity had been observed to be associated with
geological structures, which would have to be mined using
the longwall system. Thus, a system retaining the flexibility
of scattered mining, while at the same time providing
regional support in the form of stabilizing pillars, was
preferable.
Applegate
4
was the first to recommend 30 metre-wide dip
stabilizing pillars on 200 metre centres, resulting in an 85
per cent extraction ratio. Strike stabilizing pillars were also
recommended, and pillar layouts for the entire mine were
drawn up and analysed. It was recognised that these initial
layouts would be modified as new information became
available, and that detailed analysis would be carried out on
a section by section basis, taking into account mining,
geological structure, and grade considerations. Applegate and
Arnold
5
concluded that dip stabilizing pillars were more
suitable than strike stabilizing pillars at Elandsrand, if
flexibility was to be maintained in an erratic orebody. Their
reasoning appears in Table I.
By early 1990, a system of 30 metre-wide dip stabilizing
pillars on a 200 metre centre to centre spacing had been
planned for mining from 76 to 85 Levels (Applegate and
Arnold
5
). It was proposed that this spacing be retained for
long-term planning purposes, but that the final designs be
customized for each area. All major geological features would
be bracketed, while the systematic dip stabilizing pillars
would be sited in unpayable ground wherever practicable. It
was noted that accurate geological and reef grade information
would be required early so that the pillar layout design could
be finalized timeously.
Despite the choice of dip pillars, Applegate and Arnold
5
realized that there was no single correct solution to the
A review of the sequential grid mining method employed at Elandsrand Gold Mine
159 The Journal of The South African Institute of Mining and Metallurgy MAY/JUNE 2000
Figure 3Comparison of energy release rate vs percentage extraction
between several adjacent breast and updip mining longwalls (after
Cook et al.
2
)
Table I
Advantages and disadvantages of dip and strike
stabilizing pillars in scattered mining
(after Applegate and Arnold
5
)
Dip pillars: Advantages Dip pillars: Disadvantages
Holding of panels from adjacent Access haulages must pass
raises avoidedremnant formation under the pillars
therefore prevented
No gullies carried next to pillars In high grade areas there
will be a temptation to mine
beyond the pillar limit
Pillars only reduced to final
size as stoping activity ends
The major geological trend lies at
a 20 angle to the dip direction,
therefore some bracket pillars
could double as dip stabilizing pillars
Strike pillars: Advantages Strike pillars: Disadvantages
Access haulages can be sited Holing required unless adjacent
away from the pillars raises are separated by
major faults or dykes
Stabilizing effect of strike pillar Strike gulllies must be
fairly constant whereas the maintained next to and
effect of the dip pillar changes parallel to pillars on
as the final pillar position is both sides
approached
Stope access via strike
gullies vulnerable to pillar
induced strata control
problems
The pillar must be slotted at
regular intervals to allow
access crosscuts to pass
underneath
A review of the sequential grid mining method employed at Elandsrand Gold Mine
problem, and that a systematic pillar system with either dip
or strike pillars could be equally effective, each being more
advantageous than the other in certain situations. The
purpose of the pillar designs, regardless of what they were,
was to customize them to the situation at hand, with the goal
of making the design practical, easy to implement, and
profitable.
Preliminary design of sequential grid mining method
Cook et al.
2
demonstrated that an ERR peak would soon
develop with longwall faces mining on strike toward each
other, as is shown in Figure 3. Applegate
6
showed that this
ERR peak could be kept at acceptable levels by implementing
a proper sequence of mining where stope spans were kept to
a minimum. He outlined the mining strategy as a
combination of scattered mining with a system of partial
extraction. The sequence of mining was important, hence the
term sequential. Grid refers to a grid of regularly spaced
crosscuts and raises, borrowed from the scattered mining
system. McCarthy
12
coined the name sequential grid mining
in 1991.
The overall direction of sequential grid mining is outward
from the shaft on strike, moving from raise line to raise line,
toward the eastern and western boundaries of the mining
lease area, as shown in Figure 4. Because deeper levels
would be opened up later than shallower levels, the former
would lag causing the overall mining front to be V-shaped
down-dip. This is a favourable overall mining configuration.
Control of the mining sequence at each raise line is shown in
finer detail in Figure 5. The mining from each raise first
proceeds towards the shaft, so that final pillar formation
takes place at low span (thereby limiting the ERR). At this
point the mining ERR reaches a peak, but the hazard is
eliminated by the fact that mining stops at this point. Once
the pillar formation is complete, mining from the same raise
proceeds away from the shaft towards the stopping line for
the next pillar. During this phase, stope spans reach their
maximum, but this is somewhat negated by the fact that the
mining faces are advancing towards solid ground. Overall,
the ERR is well controlled, with high stresses (hence ERR)
being confined to the pillars, which are left unmined.
Raise spacings are limited by maximum practicable
scraping distance. It was clear that if bracket pillars doubled
as dip pillars, then they would not necessarily lie in the
middle between two raises as shown in Figure 6. This would
require a small raise spacing (to limit maximum scraping
distance), while pillar widths of at least 30 metres and an
extraction ratio of at least 85 per cent would require large
raise spacings. A 200 metre raise spacing was chosen as a
reasonable compromise (Applegate
6
).
Computer modelling to test the suitability of
sequential grid mining
Applegate
6
carried out computer modelling to test the
sequential grid mining concept, using MINSIM-D (Chamber of
Mines Research Organisation
7
). The modelling analysed the
sequential grid mining method, and compared it with an
alternative mini-longwall mining method in which strike
stabilizing pillars were used. The analysis was repeated in
both cases with and without backfill, and with overhand and
underhand face configurations.
In all cases, the leads and lags between faces were kept at
10 metres, and face lengths were 35 metres. The dip pillar
configuration modelled included 30 metre-wide pillars and 35
metre pillars while maintaining the same extraction ratio of
85 per cent. This was done to investigate the stability of the
dip pillars using the Hoek and Brown
8
failure criterion. All
layouts were conceptual and kept as simple as possible. The
model geometries are shown in Figure 7. The primary
variables studied in the models were the ERR on the mining
faces, and the average pillar stresses (APS) including the

160 MAY/JUNE 2000 The Journal of The South African Institute of Mining and Metallurgy
Figure 4Overall strategy for sequential grid mining
Figure 5Stoping sequence for sequential grid mining (not to scale,
after Applegate
6
)
Figure 6Example of bracket pillar replacing dip stabilizing pillar
Mining from Raise A towards dyke
done in underhand configuration
Bracket pillars on dyke
Mining
direction
from
Raise B
worst case scenario where everything has been mined except
for the stabilizing pillar system. In addition, the effect of
pillar induced stresses in the footwall were evaluated and
used to establish an optimum depth below reef that the
access haulages should be sited.
The modelling parameters employed are listed in Table II
so that Applegates
6
work can be repeated. Note that
subsequent versions of MINSIM for MICROSOFT WINDOWS
and other tabular mining analysers such as BESOL and
MAP3D should produce comparable results to those obtained
from MINSIM-D.
The modelling results are summarized as follows (after
Applegate
6
):
the average pillar stress, estimated for a depth of 2800
metres below surface at 454 MPa, was less than the
currently accepted maximum level of 2.5*UCS = 625
MPa in the case of Elandsrand (Chamber of Mines
9
)
according to the model, the APS would only exceed the
625 MPa criterion at a depth of 3860 metres, thus the
current design appears satisfactory for mining to at
least the southern boundary of the lease (about 3300
metres below surface)
dip pillar stability would therefore not be a problem
according to the above criterion, although this was
contradicted by the Hoek and Brown
8
failure criterion
(Applegate
6
)
the modelling results suggested a failed rock mass
around the pillars in both cases, when the Hoek and
Brown
8
failure criterion was employed, which might
point to future pillar instability
the planned depth of haulages at 80 metres below reef
satisfied the design criterion that a haulage should
never be subjected to a field stress exceeding 120 MPa
the ERR for both configurations was kept to acceptable
levels (20 MJ/m
2
or less) in all situations
the sequential grid mining method was able to meet
the ERR criterion by proper sequencing of the mining
the sequential grid mining system was disadvantaged
in the case where mining was carried out down-dip of
a mined out areain this case the topmost panel in the
overhand configuration gave unacceptably high ERR
values, while an underhand configuration went some
way to ameliorating this problem
the average ERR for both the sequential grid mining
system and the mini-longwall system were 16 MJ/Ca
and 13 MJ/Ca for the case of no backfill and backfill
respectivelya result not surprising as both had
similar extraction ratios
backfill reduced the APS in the sequential grid mining
system by 7.5 to 20 per cent
backfill was found to be more effective in sequential
grid mining since backfill effectiveness depends on
stope closure. Maximum stope closure occurs at the
raise line in sequential grid mining, while the highest
closure rates are found in the centre panels of the mini-
longwall system. In the former, backfill is placed in a
low closure environment (when stope spans are small),
but must be placed in a high closure environment
behind the advancing panels (in the centre of the mini-
longwall).
In general both the sequential grid mining system and
the mini-longwall system met all the rock engineering
requirements for deep level mining, except the potential for
stabilizing pillar failure, according to the Hoek and Brown
8
failure criterion. Experience at Western Deep Levels
(Hagan
10
) showed that strike stabilizing pillars on the
Carbon Leader Reef were prone to failure. Dip pillars were
expected to be more stable than strike stabilizing pillars
because they are aligned parallel to the overall ride direction,
and are more evenly loaded than strike stabilizing pillars
(Starfield and Crouch
11
). They were therefore expected to be
able to bear higher loads than strike stabilizing pillars before
becoming unstable. This suggested that dip stabilizing pillars
A review of the sequential grid mining method employed at Elandsrand Gold Mine
161 The Journal of The South African Institute of Mining and Metallurgy MAY/JUNE 2000
Figure 7Comparative MINSIM models for pillar stability in mined out
case
Table II
Modelling parameters used to evaluate the
sequential grid mining
General elastic constants Value
Poissons ratio 0.2
Youngs modulus 70.0 GPa
Stoping width 1.5 metres
Cohesion 5.0 MPa
Angle of friction 30 degrees
Backfill variables Value
Cohesion 0 MPa
Angle of friction 30 degrees
Material type Hyperbolic
Critical stress parameter (a value) 11.1 MPa
Ultimate strain parameter (b value) 0.40
Fill width 1.5 metres
Hoek and Brown failure criterion variables Value
Rockmass quality Very good
Uniaxial compressive strength of rock 200 MPa
Empirical constantm value 7.5
Empirical constants value 0.1
General computer modelling parameters Value
Coarse window grid size 20 metres
Fine window grid size 5 metres
Panel lead length 10 metres
Model depth assumed (95 level) 2700 metres
below surface
Maximum number of solution iterations 15
Successive over-relaxation factor (all iterations) 1.5
Overall stress tolerance 0.2 MPa
Maximum number of iteration cycles 15
Number of lumping shells 3
Iteration start control number 0
A review of the sequential grid mining method employed at Elandsrand Gold Mine
may not show the foundation failure characteristics
experienced with strike stabilizing pillars given the APS
figures calculated in the numerical models.
McCarthy
12
investigated the production potential of
sequential grid mining as compared with a mini-longwall
layout, with the results listed in Table III. McCarthys study
was based on the mining planned between 76 and 85 Levels,
each layout being used to access and exploit the same area of
the mine. The sequential grid system allowed a greater
amount of reef to be mined while needing less access
development. It was expected from this exercise that
sequential grid mining would be more profitable because it
both increased revenue while at the same time reduced
development costs. It was assumed that other mining costs
would remain more or less the same. The choice of sequential
grid mining for Elandsrand was therefore a calculated risk,
with clear production benefits, enhanced pillar stability,
reduced access costs, but still unknown pillar behaviour in
the long term.
The decision to implement sequential grid mining was
eventually based on the following factors:
less off-reef development: sequential grid mining
would produce 28.7 stope centares per off-reef
development metre (including boxholes), whereas the
mini-longwall system would produce 17 to 24 stope
centares per off-reef development metre, depending on
the access development layout
improved selectivity and flexibility in sequential grid
mining with respect to planning and mining, which is
essential for a variable orebody such as the VCR
recovered grade in sequential grid mining is better
because of the improved selectivity, flexibility, and
reduced off-reef mining
improved available face length in sequential grid
mining, which forms a good basis for selectivity and
flexibility
an expected reduction in seismicity in sequential grid
miningno panels would hole into one another, and
no panels would need to negotiate major geological
features
strike stabilizing pillars require ventilation slots, which
means dangerous miningsequential grid raise lines
are naturally separated into ventilation districts and no
pillar slots are required
strike gullies in sequential grid mining will in general
enjoy more stable rock conditions because none are
placed next to dip stabilizing pillars whereas the mini-
longwall system requires gullies to be placed next to
strike stabilizing pillars
backfill will be relatively more effective in the
sequential grid mining system, than in the longwall
mining system
sequential grid mining is dependent on development
staying ahead of the mining front, but this enables
accurate geological information to be gathered and
included in final designs before mining commences
this in itself leads to better planning, safer working
conditions and improved profitability.
Elandsrand therefore elected to implement the sequential
grid mining system, starting from 76 Level in early 1990.
Practical implementation of sequential grid mining
Sequential grid mining was implemented below 76 Level, but
its full impact was delayed for several years as increased
mining of the deeper reserves slowly replaced depletion of
shallower reserves above 76 Level. The practical implemen-
tation of the sequential grid mining system has taught some
lessons both from the seismic and practical mining points of
view.
Practical Mining Lessons
Figure 8 gives an overall view of how sequential grid mining
has been implemented at Elandsrand. The formation of dip
and bracket pillars is clear, especially for 1995. It can also be
seen that some stopes in the scattered mining areas above 76
Level are separated by dip and bracket pillars. The down-dip
V-shaped configuration developed between 1991 and 1995.
The development and raises ahead of the mining front are
visible in Figure 8, and these all provide advanced access to
the orebody and the timeous accumulation of accurate
geological and grade information before mining commences.
The design process employed at Elandsrand for the
sequential grid mining system is described below (after
Applegate
6
).
Preliminary design: lay out dip pillars at the standard
spacing, and adjust these where necessary if bracketing
of geological structures is planned, or patches of low
value reef have been found. At this stage designs are
based on practical situations, and rules of thumb.
Computer modelling at this stage is inappropriate

162 MAY/JUNE 2000 The Journal of The South African Institute of Mining and Metallurgy
Table III
A comparison between production from a longwall
system and sequential grid mining (after McCarthy
12
)
Mining parameter Mini-longwall Sequential grid
Square metres mined 3 623 000 3 720 000
Development metres required 219 880 157 550
Gold produced (kg) 202 535 204 297
Square metre/metre development ratio 16.9 28.7
Figure 8 Development of sequential grid mining system on Elandsrand
because of the lack of accurate geological information
in the area to mined. The rough design is suitable for
the long-term mine plan, where important aspects such
as extraction rates and development needs can be
determined
Final design: once drilling and raise mapping improve
geological information concerning grade and structure,
the preliminary designs can be customized to suit the
situation. At this stage is is essential that personnel
from rock mechanics, geology, survey and mining co-
operate to produce the final design. This may involve
re-siting dip pillars in unpayable ground, designing
bracket pillars for a dyke, etc. Once alternatives are
proposed, numerical modelling is carried out to assist
in deciding upon the best design for the area. All
personnel involved in the above process should agree
to any further adjustments to the design made from the
modelling results
Design changes: new information may force changes to
the final design, which again should be checked by
numerical modelling. The procedures listed above
should be adhered to when changes are made. Any
unauthorized changes, such as mining beyond pillar
limits must be strongly discouraged, as this could
compromise profitability and safe working conditions.
Mining commences once a reef raise holes on the next
level from the level below. When holing is complete, the reef
raise is secured by mining it from the top using a down-dip
ledging strategy. This is done by advancing a 10 metre-wide
face, angled 20 to 45 to the raise line, and scraping broken
ore down-dip into the raise. One ledging face leads the other
by a distance of not more than 10 metres. The ledged raise is
supported by a combination of packs and backfill ribs since it
has been found that the central raise line remains much more
stable during the stoping operations if it has been supported
by backfill ribs. This has been found to be true whether or
not backfilling is carried out during the stoping phase.
Typically, a raise takes 8 months to hole from one level to
the next, and down-dip ledging takes another 8 months
between levels. It takes roughly two years to mine out a raise
line between two levels, i.e. replacement of mined-out raises
takes roughly two-thirds the time it takes to mine out a raise
line between levels. Once stoping of a raise line is completed,
sweeping and vamping operations are carried out before the
crosscut is closed at the breakaway position. These time
constraints dictate that at least three raise lines are in various
stages of development ahead of the main mining front on
each level. The crosscut nearest the haulage end should be
broken away from the haulage, the second developing
towards reef, and the third raising on reef. Mining will be in
progress on the fourth raise line from the haulage end (i.e.
about 600 metres behind the haulage development).
A delicate balance must therefore be struck, where
development is kept sufficiently far ahead of mining so that
newly ledged raises become available for stoping when
required by the sequence. It has been found that if
development gets too far ahead, ledged raises stand for too
long a period before being mined (in some exceptional cases,
two years), and strata control problems develop in the raise
and crosscut before mining commences. In terms of
selectivity and flexibility, the further the development is
ahead of mining, the better. The three raise-line compromise
on each level appears to have worked best so far, and is used
as a measure to plan future development requirements.
Presently, ledging contributes about 25 per cent of the reef
centares produced by the mine.
Sequencing of stoping operations exactly as proposed in
theory is extremely difficult to achieve in practice. As a
result, there have been and will be many instances where the
theoretical ideal cannot be followed for various reasons, e.g.
the presence of poor grade, unknown geology, or
unexpectedly severe seismicity. These conditions could lead
to unplanned delays and re-planning. One consequence is
that panels from two adjacent raise lines could end up mining
toward each other. It was observed that seismicity in the
planned pillar area between the two advancing stopes would
increase dramatically before the panels had reached the
planned stopping positions, thereby increasing the potential
for rockburst accidents. This had occurred on several
occasions relatively early, and rules to control it were
required. A study by Handley
13
revealed that mining-induced
stress in the planned pillar area increased suddenly when the
panels were 70 metres apart on strike, i.e. when both sets of
panels approaching the pillar position were approximately 20
metres from their stopping positions.
The model showed that if one set of panels was stopped,
the rate of stress increase in the pillar area was reduced, and
by observation, so was the seismicity. The other set of panels
could then mine in relative safety to the pillar stopping
position. Once this had been achieved, the second set of
panels could resume mining, and also reach the pillar
stopping position in relative safety. This became known as
the 70 metre rule, and is still applied at Elandsrand today.
Essrich
14
has compiled a report covering problems and
their remedies in sequential grid mining. These are listed
with the appropriate remedy in Table IV.
These problems have all arisen in the past ten years of
mining, and have been solved by the remedies given in Table
IV. Overall, sequential grid mining can be safely and
profitably applied in the VCR environment at depths similar
to those at Elandsrand. Furthermore, the rock-related
problems encountered using this method can be solved by
existing knowledge and experience.
Performance of sequential grid mining at Elandsrand
Performance measures for sequential grid mining considered
in this paper are seismicity, productivity, and rock-related
injury rates. The discussion below is broken up into these
three categories.
Seismicity
Before 1992, an analogue seismic location system was
functioning, but the data it produced was both inadequate
and unreliable. In 1992, a digital seismic system produced by
Integrated Seismic Systems International of Welkom, South
Africa (ISSI) was installed underground. This system is a
fully digital real-time seismic and non-seismic geotechnical
monitoring system. It provides facilities for on-line quanti-
tative data processing, data interpretation and data visual-
ization. The seismic system consists of the following
equipment:
A review of the sequential grid mining method employed at Elandsrand Gold Mine
163 The Journal of The South African Institute of Mining and Metallurgy MAY/JUNE 2000
geophones installed underground and on surface
underground data acquisition and communications
equipment including intelligent, processing and multi-
seismometers
electrical cable linking underground equipment with
surface computers
a central control site with computers
processing, interpretation and visualisation software.
All the seismic hardware is modular in design, making it
easy to add and remove components. This is particularly
useful when extending the system as the mine expands to
exploit new areas. Deelkraal possesses the same type of
system, which became functional in 1995.
A count of mine seismic frequency for all seismic events
greater than or equal to magnitude 1.0 on the local
Magnitude Scale have been plotted in Figure 9 against the
build-up of sequential grid mining at Elandsrand. This graph
clearly shows a major drop-off in seismic activity induced by
mining during 1993 and 1994, after which it has remained at
20 to 30 per cent of former levels. This cannot be ascribed to
sequential grid mining alone, but the introduction of the new
mining method must have played a part. Other factors which
played a role are discussed in Section 4 above. Although
Figure 9 is compelling evidence of the success of sequential
grid mining, a more rigorous analysis of seismic activity was
still considered necessary to establish it as a viable deep level
mining strategy. Comparative analyses carried out by
SIMRAC
15
and Essrich
16
have been done, and are described
below.
SIMRAC
15
have compared the seismic response to
longwall mining with that of sequential grid mining by
considering Elandsrand and Deelkraal, both of which are
mining VCR in a similar geotechnical environment at similar
depth. Variables such as geology, rock type, stoping width,
and depth can be ignored for the seismic comparison, and
differences can thus be ascribed to the mining method. The
seismic data from both mines was recorded in 1996, when
both operated similar seismic location systems (the type of
seismic system could have an effect on the seismic statistics
obtained).
The results quoted by SIMRAC
15
have been reworked
slightly and presented in Table V. The number of seismic
events has been normalized to the area mined on each mine,
to illustrate the relative seismic frequencies encountered at
each mine. It can be seen that for magnitude 0.8 events and
larger, approximately 45 per cent more can be mined using
the sequential mining system than is the case for the
longwall system. For events with magnitude exceeding 2.0
(which have a good chance of damaging underground
excavations), approximately 153 per cent more can be mined
with the sequential grid system than with the longwall
system. This means that the seismic hazard due to
potentially damaging events is far lower in a sequential grid
mine than in an equivalent longwall mine.
SIMRAC
15
quantified the seismic hazard for each mining
method using the variables V
eff
and VGM (see Table V). V
eff
is the volume of rock in which all the seismic events
occurredi.e. a measure of clustering. For example, a large
number of seismic events clustering into a small volume
would pose a threat to any mining excavation in that volume,
because their potential to damage it would be high. If the
seismicity is more diffuse (i.e. has a larger V
eff
) then the
damage potential to an excavation in the volume would be
less. VGM, the volume of ground motion, integrates standard
seismic hazard parameters (b-value, seismic activity rate,
and maximum magnitude) into one parameter that combines
the statistical and physical attributes of seismicity with its
damaging potential (Mendecki
17
). A smaller VGM means that
a relatively smaller volume of rock has been affected by
damaging seismic events i.e. the damaging potential of all the
seismicity recorded is likely to be less.
Normalizing for the two mining methods given in Table
V, VGM/V
eff
for the sequential grid mining system is
0.85/1.80 = 0.47, and for the longwall system 1.10/0.50 =
2.20. Finding the ratio between these numbers i.e. 2.20/0.47
= 4.68, means that the longwall system can be considered to
be 4 to 5 times more dangerous than sequential grid
mining, based on the seismic statistics analysed (SIMRAC
15
).
Table IV
Problems and remedies associated with sequential
grid mining (after Essrich
14
)
Problem Remedy
Minor pillar seismicity encountered Wider pillars and smaller
up to two raise lines behind stope spans at greater depths
current mining
Seismicity on geological structures Appropriate bracket pillar
planning and designs for all
active geological structures
Geological structure intersecting Avoid placing haulage in geological
haulage under pillar feature directly under pillar
Difficult to mine top panel in Apply underhand configuration
overhand configuration in in abutments below
abutment (see points 1 and 2 in mined out areas
Figure 4)
One-sided loading of abutment Start up one or two panels, mining
while mining first side of ledge slowly to allow rockmass to
release stress
Starting panels from ledge, or Ensure stiff last line of face support,
re-starting stopped panels increase support density during
start-up phase
Lack of flexibility Correct long-term planning
Large seismic events in mining Reduce number of active panels
final stages of raise line section and mined panel length towards
end of raise life
Remnant pillar mining Avoid totally unless alternative
regional support system is in
place (e.g. backfill)
A review of the sequential grid mining method employed at Elandsrand Gold Mine

164 MAY/JUNE 2000 The Journal of The South African Institute of Mining and Metallurgy
Figure 9Production statistics for Elandsrand from 1988 to 1997
One can also look at the relative seismic hazard between
the two mining systems by again considering VGM in Table V
above and the area mined (assuming that the stoping widths
are the same). Normalizing the VGM per unit area mined (in
this case choose the unit area mined as 100 000 ca) for
sequential grid mining gives 0.85/3.18 = 0.27, and for
longwall mining 1.10/1.97 = 0.56. This means that in terms
of area mined, longwalling is more than twice as dangerous
(0.56/0.27 = 2.07) from a seismic hazard point of view as
sequential grid mining (SIMRAC
15
). It must be noted that the
data presented here was gathered over one year only, and
conclusions may change for larger data sets. In summary, it
appears that sequential grid mining is the safer method for
the VCR in the western part of the Carletonville Goldfield, but
that larger data sets are necessary to define the relative
seismic hazard more precisely.
Essrich
16
carried out a similar comparison between the
76-36 Longwall West at Western Deep Levels and the 80-32
W2 to W5 Panels at Elandsrand (Table VI). The latter panels
formed part of the sequential grid system at Elandsrand. In
both cases, the stopes were bounded by dykes to the west; at
Elandsrand a bracket pillar was left on the Footwall Dyke
when the panels reached their planned mining limit, and at
Western Deep Levels the toe of the longwall was stopped on
the Georgette Dyke. The different methods used to approach
the geological structures at the two sites are likely to be
responsible for some of the differences in the measured
seismic responses. While the intersection of the South
Georgette and Tarentaal Dykes in the case of the longwall
was a centre of high seismic energy release during mining,
no such concentration of failures was observed on the
Footwall Dyke at Elandsrand. Instead, the seismic energy
release in the sequential grid stope showed a positive
correlation with area mined, and showed a peak in the 4th
quarter of mining, and a decline afterwards. When the
mining approached closest to the Footwall Dyke, seismic
energy release had already dropped to pre-peak levels. This
study shows that the integration of major geological features
into bracket pillars in the sequential grid mining environment
plays an important role in reducing the seismic hazard,
because mining is not required to intersect the dyke, as in the
case of the longwall.
The measurements listed in Table VI were not the same
as those of SIMRAC
15
, but some of the results are very
similar. It can be seen that the mining in the longwall was
much more intensive, with much higher production rates,
from essentially the same volume of rock. This is the most
important driving force behind mining-induced seismicity,
and probably accounts for the differences in seismic rates
(expressed as an average number of centares mined per
seismic event) presented in Table VI. For example, the
longwall experienced much higher seismic rates than the
sequential grid stopes, all compressed into a period of five
months for the former, compared with the 18 months for the
latter. The impact this may have had on safety is not known
since Essrich
16
confines his discussion to mining-induced
seismicity.
Comparing the seismic rates in Tables V and VI it is clear
that the results are similar. It would appear that both studies
typify the differences between longwall and sequential grid
mining, since both were carried out at different times, and
compared different mining areas. It is concluded, from the
findings of SIMRAC
15
and Essrich
16
, that sequential grid
mining is safer than longwall mining on the VCR in the
western part of the Carletonville Goldfield.
Productivity
Sequential grid mining should result in productivity
improvements; these are not necessarily a direct consequence
of the mining method, and could also arise from good
management practice. Therefore only two measures are
considered, namely area mined annually and average area
mined per linear metre of development. Both measures are
plotted in Figure 9 with the mining-induced seismicity.
A review of the sequential grid mining method employed at Elandsrand Gold Mine
165 The Journal of The South African Institute of Mining and Metallurgy MAY/JUNE 2000
Table VI
Comparison of seismicity generated by a longwall
and a sequential grid stope (after Essrich
16
)
Parameter Sequential Longwall Ratio
Grid (SG) (LW) (SG/LW)
Reef type VCR VCR -
Period over which mining Sep. 1994 May 1996 3.60
took place to Mar. 1996 to Sep. 1996
(18 months) (5 months)
No. of panels mined 4 9 0.44
Block dimensions in which mining 140 x 120 360 x 40 -
took place (metres)
Average face advance rate 6 7 0.86
(metres per month)
Average depth of mining 2200 2100 1.05
(metres below surface)
Total area mined (square metres) 12500 11800 1.06
Average area mined per month 700 2360 3.37
(square metres)
Average sq. m mined per event 431 (29) 223 (53) 1.93
with mag. 1.0 (total no. events)
Average sq. m mined per event 12500 (1) 2360 (5) 5.30
with mag. 2.0 (total no. events)
Cumulative seismic energy 304 473 0.64
measured (MJ)
Area backfilled expressed as a 0 20 -
percentage of area mined
Rockburst related accidents 2 reportable 2disabling -
5 reportable
Table V
Comparison of seismicity generated by a longwall
system and sequential grid mining at Elandsrand
(after SIMRAC
15
)
Parameter Sequential Longwall Ratio
Grid (SG) (LW) (SG/LW)
Area mined (centares) 318357 197402 1.61
Ave. sq. m mined per event with 428 (743) 296 (667) 1.45
mag. 0.8 (total no. events)
Ave. sq. m mined per event with 22740 (14) 8973 (22) 2.53
mag. 2.0 (total no. events)
Magnitude of largest event 3.0 2.9 1.03
Magnitude of second largest event 2.6 2.8 0.93
V
eff
effective volume covered by 1.8 0.5 3.60
seismic activity (km
3
)
VGMcumulative volume of rock 0.85 1.1 0.77
affected by ground motion 55
mm/s (km
3
)
A review of the sequential grid mining method employed at Elandsrand Gold Mine
There has been a steady increase in the area mined on
the VCR annually, from 392 000 square metres in 1988 to
486 000 square metres in 1997, a 24 per cent increase over
the period (Figure 9). This is not remarkable in itself until
the seismic statistics are considered: in the same period there
was an approximate 80 per cent drop in mining-induced
seismicity (Figure 9). When considering the comparisons in
the previous section between longwall mining and sequential
grid mining, it would appear that sequential grid mining
meets expectations for the control of mining-induced
seismicity. The curves in Figure 9 therefore show that it is
possible to manage mining-induced seismicity effectively
using the sequential grid mining system.
During the same period sequential grid mining has
allowed the stoping to development ratio to improve by 22
per cent from 18.50 square metres per linear metre developed
to 22.55 square metres per linear metre (Figure 9). This
improvement falls short of McCarthys projected figure of
28.7 square metres per linear metre (Table III) by 21 per
cent, but is still a 33 per cent improvement on the
comparative figures expected for the mini-longwall layout.
The reason why sequential grid mining has not reached
projected expectations is that the mine needed a far greater
development rate below 76 Level to open the orebody
sufficiently for the expected mining flexibility. Therefore
annual development rates have equalled or exceeded the
minimum required to regenerate mined out reserves using
the sequential grid method. This rate should flatten in future
while stoping rates will continue to increase slightly, which
should result in further improved stoping to development
ratios.
Rock related injury statistics
Rock related injury statistics (including both rockbursts and
rockfalls) can be used to reflect the anticipated improvements
offered by sequential grid mining. For this purpose, the work
by Moalahi
18
has been summarized in this section and
presented in Figure 10. The statistics cover a period of 10
years (January 1988 to December 1997), spanning the time
that sequential grid mining was introduced and implemented
on Elandsrand. The rockfall and rockburst injury rates were
calculated by counting the total number of persons injured
(both fatally and non-fatally) each year and expressing the
result as a rate per 100 000 centares mined.
Plots 1 and 3 in Figure 10 compare rockfall injury rates
for the scattered mining and sequential grid mining methods
at Elandsrand. Application of the sequential grid mining
method is not expected to influence rockfalls, which should
be controlled by other factors. The recent downward trends
for both scattered mining and sequential grid mining are
similar, showing that an amalgam of mine-wide factors
contributed to the reduction in rockfall injury rates e.g.:
improved support methods (Hamman
19
)
improved safety management including for example
training, barring down, and compliance with support
recommendations
recognition of the influence of slope reef conditions
on the stability of the hangingwall in stopes (Hamman
and Close
20
).
The curves in plots 1 and 3 help to confirm that the
introduction of sequential grid mining ought not to influence
rockfall injury rates.

166 MAY/JUNE 2000 The Journal of The South African Institute of Mining and Metallurgy
Figure 10Rockfall and rockburst injury statistics for Elandsrand from 1988 to 1997 (after Moalahi
18
)
The rockburst injury rates compare the benefits of
scattered mining and sequential grid mining in plots 2 and 4
in Figure 10. In plot 2 reductions in overall rockburst injury
rates have been larger in the scattered mining area than in
the sequential grid mining area. It appears that support
improvements, insights, safety management, the application
of the remedies listed in Table IV, and other factors
mentioned above brought about reductions in rockburst
injury rates during the past five years. The results for the
scattered mining area could have been helped by a reduction
in mining rate during the same period, and the fact that the
average mining depth is 500 metres (18 per cent) less than
the sequential grid mining area.
Strong downward trends in the rockburst fatality rate are
present for both scattered and sequential grid mining
between the beginning of 1993 and the end of 1997 (Figure
10, plot 4). Here, improved support and the recognition of
unstable hangingwall (Hamman
19
and Hamman and Close
20
)
would have played particularly important roles. As a result,
there has been a 100 per cent reduction in rockburst fatalities
from 1992 to 1995 in the scattered mining area, and no
fatalities since. From 1992 to 1997, there has been an
approximately 80 per cent reduction in the rockburst fatality
rate in sequential grid mining, which coincides with the
reduction in mining induced seismicity shown in Figure 9.
Improved support, recognition of unstable hangingwall
conditions, and the application of remedies listed in Table IV
would have contributed to this reduction, but it is expected
that reduced seismicity could have played a leading role. The
linear correlation coefficient between mining-induced
seismicity and the rockburst fatality rate was calculated to be
0.69. It is proposed that a detailed study into the causes of
rockburst injury statistics be undertaken for sequential grid
mining in order to identify the relative benefits that different
factors such as support, etc. have contributed to safety and
productivity.
In summary, Elandsrand has reduced all rock-related
injury rates significantly in the past five years. This has been
the result of several improvements, and cannot be ascribed to
sequential grid mining alone. The flexibility inherent in
sequential grid mining may have contributed to the reduction
in mining-induced seismicity, which in turn is likely to have
reduced rockburst injury rates. Again, other factors would
also have contributed to this reduction. A more detailed
study of rock related injury rates is necessary to identify the
contributions made by the different factors.
Conclusions
It is concluded that, although the evidence presented is not
complete, sequential grid mining is a simple methodical way
of exploiting the deep level Ventersdorp Contact Reef reserves
at Elandsrand which:
results in a regional dip pillar support system
keeps energy release rates to an acceptable minimum
by sequencing mining in an optimal manner on a
regular grid system
allows pillar location on geological features and low
grade areas, thereby maximizing gold recovery
provides good geological and reef grade information for
short- and long-term mine planning and design
purposes
offers the opportunity of revising mining plans before
mining commences when new geological information
comes to light
requires less linear development per square metre
stoped than an equivalent mini-longwall system
through its flexibility of application enables effective
management of mining-induced seismicity and
associated losses
has to date not presented the miner with any insoluble
mining or ground control problems
by comparison with longwall mining methods applied
at similar depths in the western part of the Carletonville
Goldfield, has been shown to be the safer method both
in terms of mining-induced seismicity and rockburst
injuries.
There are four negative aspects to sequential grid mining,
namely:
the regular mining layout can result in lack of
flexibility in planning and mining
sequencing the mining limits the amount of face
available for exploitation
sequential grid mining requires a higher up-front
capital input to establish mining faces than does
longwall mining, which in turn can have a negative
impact on the internal rate of return on a mining
project
the long-term potential for instability in the pillars is
still unknown.
There are, as yet, no viable solutions for the first three
problems. In terms of the fourth problem, no disruptive pillar
failures have occurred to date, but the potential remains, and
will increase as more pillars are formed. The above-named
problems need to be solved by further research before
sequential grid mining can be unequivocally claimed to be
the preferred method of extracting deep level tabular
orebodies. Sequential grid mining has been shown not to
influence rockfalls, but it ought to influence rockbursts, and
more work needs to be done to establish the magnitude of
this effect.
To date, experiences with sequential grid mining at
Elandsrand have shown that it is possible to turn a deep,
medium grade orebody with erratically distributed gold
values to profitable account.
References
1. ENGELBRECHT, C.J. The West Wits Line. In : Antrobus E.S.A. (Ed.),
Witwatersrand Gold100 Years, Geological Society of South Africa,
Johannesburg, 1986.
2. COOK, N.G.W., HOEK, E., PRETORIUS, J.P.G., ORTLEPP, W.D., and SALAMON,
M.D.G. Rock mechanics applied to the study of rock bursts. J. S. Afr. Inst.
Min. Metall. vol. 66. 1966. pp. 435528.
3. MURIE, A. The Regional Effects of Leaving Unpay Blocks of Ground and/or
a System of Regularly Spaced Rib Pillars for the Scattered Mining Areas
above 2238 Level. Internal Memorandum, Elandsrand Mine, AngloGold
West Wits Operations, 29th October 1980.
4. APPLEGATE, J.D. Preliminary Comments on Stabilising Pillars. Internal
Memorandum, Elandsrand Mine, AngloGold West Wits Operations,
Reference 572/EGM171/86M, 14th July 1986.
5. APPLEGATE, J.D. and ARNOLD, D.A. Stabilising Pillar Design - 76 to 85
Levels. Internal Report, Elandsrand Mine, AngloGold West Wits
Operations, Reference 572/EGM1/017/90R, 2nd March, 1990.
6. APPLEGATE, J.D. Rock Mechanics Aspects of Sequential Grid Mining. Project
Report submitted to the University of the Witwatersrand in partial
A review of the sequential grid mining method employed at Elandsrand Gold Mine
167 The Journal of The South African Institute of Mining and Metallurgy MAY/JUNE 2000
A review of the sequential grid mining method employed at Elandsrand Gold Mine
fulfilment of the Degree of Master of Science in Engineering,
Johannesburg, 1991.
7. Chamber of Mines Research Organisation (Now CSIR Division of Mining
Technology). MINSIM-D Users Guide. CSIR Division of Mining
Technology, Johannesburg, 1985.
8. Hoek, E. and Brown, E.T. Underground Excavations in Rock. The
Institution of Mining and Metallurgy, London, 1980.
9. Chamber of Mines of South Africa. An Industry Guide to Methods of
Ameliorating the Hazards of Rockfalls and Rockbursts - 1988 Edition.
Johannesburg : The Chamber of Mines of South Africa, 1988. 114 pp.
10. HAGAN, T.O. Pillar foundation failure studies in a deep South African gold
mine. In: Fairhurst, C. (ed.). Rockbursts and Seismicity in Mines,
Proceedings of the Second Symposium on Rockbursts and Seismicity in
Mines, Minneapolis, June 1988. Rotterdam: A.A. Balkema, 1990. pp.
6569.
11. STARFIELD, A.M. and CROUCH, S.L. Elastic analysis of single seam
extraction. Proceedings of the 14th U.S. Symposium on Rock Mechanics,
American Society of Civil Engineers, 1973.
12. MCCARTHY, T.F. Production Aspects of Sequential Grid Mining.
Unpublished presentation material, Elandsrand Mine, AngloGold West
Wits Operations, 1990.
13. HANDLEY, M.F. Internal Memorandum, AngloGold West Wits Operations,
18 May 1993.
14. ESSRICH, F. Review of Seismicity in Sequential Grid Mining on Elandsrand.
Internal Report, Reference No. 012/98, Elandsrand Mine, AngloGold West
Wits Operations, 3 February, 1998.
15. SIMRAC: Mine Layout, Geological Features and Seismic Hazard. Mid-year
report, GAP 303, Department of Mineral and Energy Affairs, Pretoria, June
1997.
16. ESSRICH, F. Comparison between Seismicity Associated with the Mining of
the 80/32 Raise Line (EGM) and the 76/36 W Longwall (WDL West).
Internal Report, Elandsrand Mine, AngloGold West Wits Operations,
Reference no. 42/96, 29 October 1996.
17. MENDECKI, A.J. An Attempt to Estimate Seismic Hazard on the AAC Gold
Mines. AAC Internal Report Reference SP8/85, 1985.
18. MOALAHI, N.C.: Msc. thesis in preparation, Department of Mining
Engineering, University of the Witwatersrand, Johannesburg, 1998.
19. HAMMAN, J.P.E. A preliminary assessment of the performance of pre-
stressable and yielding elongates as a stope support system for
Elandsrand Gold Mine. 1st Southern African Rock Engineering
Symposium, Johannesburg, South Africa, 15-17 September 1997. pp.
228233.
20. HAMMAN, J.P.E. and CLOSE, A.J. Hangingwall conditions and its effect on
safety at Elandsrand Gold Mine. Anglo American Corporation of South
Africa Limited: Managing Directors Safety and Health Seminar, Western
Deep Levels, October 1993, pp. u245u256.

168 MAY/JUNE 2000 The Journal of The South African Institute of Mining and Metallurgy